On 15th January 2019, MMP has been selected in the Call for Growth by Next Energy (Cariplo Factory) to access the Engage Phase and start a collaborative service development and deploy test with TERNA.
The collaborative project deals with the analysis and the study of the forecast error models used by TERNA concerning the Italian electric load and energy production from wind and solar sources. The analysis aims at implementing a clustering method based on a Bayesian Gaussian Mixture Model in order to produce a non-normal asymmetric error distribution for a certain variable (load, wind or pv), market zone, hour and day. This will provide an alternative method for TERNA to estimate variable errors and then to size reserve capacity and resources during the dispatching services market (MSD).
Error distributions, which are highly asymmetric, are decomposed with a BGMM. Below some example of the methodology implemented for the different variables.
High correlations exist among different variables (mainly between the load and the solar production) and different zones for the same variable. Here some examples: bivariate Bayesian Gaussian Mixture Model is applied for the pv-load correlations for the NORD and SICI market zones and for the correlation of wind between CSUD and SUD.
